What decides this category
Whether the tool has a real model of your data, and which warehouses it connects to. The plain language querying that gets demonstrated is the easy half; the hard half is the semantic layer that makes the answer correct rather than merely plausible.
Why this is not filed under AI Code
Both write code, which is exactly why the line needs drawing. A coding assistant is bought to help a person write anything, and it will happily write you a query. A tool in this category is bought so that a question about your own data comes back correct rather than merely well formed, and that difference is the semantic layer described above rather than the model behind it.
The coding tools are judged on what they do to a repository. These are judged on what they know about a warehouse.
What we publish here
Profiles only, for now.